Jean-Frédéric Ferté’s Post

Got some spare time... Ai generated concept art for a survival/sim game. Some extended prisonner dilema story. ...on going design/tech design + narrative tone/content and writing. Let's see how fast we can progress with modern tools.

"We select a possibility and we walk until we reach it. So, in a sense, we create it." Corwin 😉

Choose a starting page in a specific 'domain' (here wikipedia, en) the toy creates a new local cache and let you browse within the defined domain, autmatically scrapping the pages (a little rule system exists for filtering the scrapped content) and keeping a diary of the page visited and their date/moment of first visit. ...multiple uses possible for education be it wetware or silicon based. ...maybe it is a hunter gatherer's information's net/bag, letting you grab the valuable information and patterns around to bring them back to your peers ?

Yop, as I needed to update (ehrm... let's say I kinda left it unattended for a decade) my personal website, I thought I could use ai assistance (Ai as tools is my new hammer, everything resemble a nail). ...it once more went beyond my expectations and was extremely instructive when it comes to game design and collective creation tasks with Ai in fact. (I should write an actual serious paper about this part !) I used another custom assistant to create the report. (...realises he is using a (crappy) self developped AI assistant to file reports on video game dev assisted by AI, ...brain explodes.)

The most challenging part with AI is to identify an optimal path using it... ...without having to reconsider this path every step as the possibilities outmatches your inital understandings by their cumulative effect. (random #slop for illustration)

app on demand... since tool creation is extremely cheap with AI, let's test building tools on demand. Example : html editor/viewer (3 prompts required, total worktime : 12 minutes) then a good agent should be able to do this proactively... (turns out I was waiting for this level of tool complexity to interactively build the design, the structure and the production pipeline in one consistent space tools and trades... 🤔 )

AI assisted post writing (with the exception of these lines, the post should read 'conventional' 😅 ). Mostly common sense advices, but it highlights the links between : competence division, communication frequence ('iteration step size') -> design/project strategy "As developers, we're always looking for ways to boost productivity. AI assistants are game-changers, but how do we effectively use them for projects too large for their conversation memory? Here are some strategies I've found effective: 1️⃣ Project Segmentation: Break your project into smaller, independent modules. 2️⃣ Structured Documentation: Create reference docs for each module. 3️⃣ Minimal Context: Provide only necessary context for each task. 4️⃣ Targeted Sessions: Focus each interaction on specific aspects. 5️⃣ Short Iterations: Work in small increments, frequently integrating generated code. 6️⃣ Code Snippets: Use snippets to illustrate project-specific patterns. 7️⃣ Targeted Questions: Ask precise questions for more relevant responses. 8️⃣ Manual Review: Always review AI-generated code before integration. 9️⃣ Decision Log: Keep a journal of major decisions and changes. 🔟 Predefined Templates: Create templates for different code parts. 1️⃣1️⃣ Version Control: Use Git, creating branches for each feature. 1️⃣2️⃣ Feedback Loop: Note AI strengths and weaknesses, adapting your approach. 1️⃣3️⃣ Tool Integration: Combine with linting and automated testing tools. 1️⃣4️⃣ Recap Sessions: Periodically review progress and challenges with the AI.By implementing these strategies, we can harness the power of AI assistants while maintaining code quality and project coherence. It's about finding the right balance between AI capabilities and human oversight. What's your experience with AI coding assistants? Any tips to add? Let's discuss in the comments!" generated with AI (slop slop slop) #AIinCoding #DeveloperProductivity #TechInnovation #SoftwareDevelopment #AICollaboration#Slop

Another feedback from homebrewing. (AI agent, basic setup using ollama and langchain. I used a 'tale teller monk personality definition and some spices, got some interesting outputs so here is one I felt suitable for sharing) ps: have the best Sunday ever !